Draft version November 18, 2016Preprint typeset using LATEX style AASTeX6 v. 1.0

RADIO-INTERFEROMETRIC MONITORING OF FRB 131104:

A COINCIDENT AGN FLARE, BUT NO EVIDENCE FOR A COSMIC FIREBALL

R. M. Shannon1,2,4 and V. Ravi3,4

1CSIRO Astronomy and Space Science, Australia Telescope National Facility, Box 76, Epping NSW 1710, Australia2International Centre for Radio Astronomy Research, Curtin University, Bentley WA 6102, Australia3Cahill Center for Astronomy and Astrophysics, MC 249-17, California Institute of Technology, Pasadena, CA 91125, USA4The authors contributed equally to this work

ABSTRACT

The localization of fast radio bursts (FRBs) has been hindered by the poor angular resolution of the

detection observations and inconclusive identification of transient or variable counterparts. Recently

a γ-ray pulse of 380 s duration has been associated with the fast radio burst FRB 131104. We report

on radio-continuum imaging observations of the original localization region of the FRB, beginning

three days after the event and comprising 25 epochs over 2.5 yr. Besides probabilistic arguments that

suggest that the association between the γ-ray transient and the FRB is not compelling, we provide

upper limits on a putative radio afterglow of this transient that are at odds with standard models

for its progenitor. We further report the discovery of an unusual variable radio source spatially and

temporally coincident with FRB 131104, but not spatially coincident with the γ-ray event. The radio

variable flares by a factor of 3 above its long term average within 10 d of the FRB at 7.5 GHz, with a

factor-of-2 increase at 5.5 GHz. Within our observations, the variable has persisted with only modest

modulation and never approached the flux density observed in the days after the FRB. We identify an

optical counterpart to the variable. Optical and infrared photometry, and deep optical spectroscopy,

suggest that the object is a narrow-line radio AGN.

Keywords: fast radio bursts — radio continuum: general — gamma rays: general — galaxies: active

— black hole physics

1. INTRODUCTION

Fast radio bursts (FRBs) represent a new phenomeno-

logical class of astrophysical transient. They are bright

(&Jy peak flux density) pulses of radio emission that

show the effects of propagating though large column

densities of plasma: dispersion through ionized plasma,

multipath propagation due to inhomogeneities in the

plasma (Lorimer et al. 2007; Thornton et al. 2013), and

Faraday rotation due to magnetization of the plasma

(Masui et al. 2015). The column densities exceed pre-

dictions for the Galaxy, suggesting that the FRBs are

extragalactic and possibly cosmological in origin (Ravi

et al. 2016). They have hitherto only been detected us-

ing single-dish telescopes, which have poor angular reso-

lution. Only one FRB has been found to repeat (Spitler

et al. 2016), greatly enhancing prospects for its localiza-

tion in follow-up observations. For the remaning FRBs,

which have not repeated, attempts at localization have

ryan.shannon@csiro.au,vikram@caltech.edu

relied on detecting counterpart multi-wavelength tran-

sients that might be expected if the FRBs arise from cat-

aclysmic explosions or outbursts. A claimed association

of FRB 150418 with a centimeter-wavelength afterglow

and host galaxy (Keane et al. 2016) has been disputed

and instead attributed to common AGN variability, ei-

ther intrinsic (Williams & Berger 2016; Vedantham et al.

2016a), or caused by Milky Way scintillation (Akiyama

& Johnson 2016; Johnston et al. 2016). With so little

detail on the locations of FRBs, theories for their pro-

duction and sources are understandably varied, ranging

from ultabright pulses from pulsars (Cordes & Wasser-

man 2016), to cosmic strings (Cai et al. 2012).

We detected FRB 131104 (Ravi et al. 2015) with the

64-m Parkes radio telescope in the direction of the Ca-

rina dwarf spheroidal galaxy (Car dSph), 100 kpc dis-

tant from Earth. The FRB has an electron column den-

sity, measured in units of dispersion measure (DM), of

779.0 ± 0.2 pc cm−3 and shows evidence for temporal

broadening associated with multipath propagation. De-

spite its detection in a targeted observation of the Car

dSph, we have no evidence to associate the FRB with

arX

iv:1

611.

0558

0v1

[as

tro-

ph.H

E]

17

Nov

201

6

2 R. M. Shannon & V. Ravi

that galaxy. The FRB has not repeated in ≈ 100 hr of

follow-up observations at Parkes.

A γ-ray transient, Swift J0644.5−5111, has recently

been associated with the FRB at the 3.2σ to 3.4σ con-

fidence level (DeLaunay et al. 2016). The emission was

detected in an off-axis position with the Swift satellite’s

Burst Alert Telescope (BAT; Barthelmy et al. 2005),

coincident with the FRB in position and time. The

transient duration was ∼ 380 s, with an inferred en-

ergy output of 5 × 1051 erg. DeLaunay et al. (2016)

suggested that the γ-ray emission (assumed to be as-

sociated with this FRB) was generated by shocked rel-

ativistic plasma in a cosmological explosion, or in an

accretion episode associated with a supermassive black

hole. We discuss the claimed association between Swift

J0644.5−5111 and FRB 131104 in Section 2, address-

ing specifically the mismatch (noted by DeLaunay et

al.) between the low rate of such transients observed by

Swift/BAT and the high FRB all-sky rate.

Here we report on a centimeter-wavelength radio mon-

itoring campaign of the Parkes localization region of

FRB 131104, and the discovery of an unusual, variable

radio source (AT J0642.9−5118) that flares coincident in

time and location with FRB 131104. AT J0642.9−5118

is not coincident with Swift J0644.5−5111; indeed, our

observations exclude any bright radio afterglow of Swift

J0644.5−5111. In Section 3, we present radio observa-

tions of the field and the light curve of AT J0642.9−5118.

In Section 4, we present an optical characterization of

AT J0642.9−5118. We discuss the implications of our

observations in Section 5, and conclude the paper in

Section 6.

2. THE γ-RAY TRANSIENT COINCIDENT WITH

FRB 131104

Swift J0644.5−5111 was discovered within the 15′ di-

ameter half-power circle of the beam (#5) of the Parkes

21-cm multibeam receiver in which FRB 131104 was de-

tected, 6.2′ from the beam center. Its position at the

edge of the BAT field of view, illuminating only 2.9%

of detectors, resulted in a 4.2σ detection in the image

plane despite its high fluence of 4 × 10−6 erg cm−2. As-

suming a distance of 3.2 Gpc for FRB 131104 based on

comparing its extragalactic DM with models for the ion-

ized content of the Universe, the isotropic energy output

of Swift J0644.5−5111 was 5 × 1051 erg, with a dura-

tion of ∼ 380 s. This is somewhat longer, and some-

what less energetic than most long-duration gamma-ray

bursts (GRBs) detected by Swift (Gehrels et al. 2009),

but is inconsistent with other GRB subtypes (e.g., ultra-

long GRBs). Murase et al. (2016) consider it likely that,

largely independent of the source model, a radio after-

glow would have been present. We constrain such an

afterglow in Section 5.2 using our observations.

DeLaunay et al. (2016) estimate a significance

for the association between FRB 131104 and Swift

J0644.5−5111 of between 3.2σ and 3.4σ, based on the

estimated false positive rate in a large collation of BAT

archival data. This corresponds to an odds ratio of be-

tween ≈ 600:1 and 1800:1. Following an argument made

by Vedantham et al. (2016a), we compare this odds ratio

with the expected number of FRBs that exhibit similar

counterparts, which we can estimate by comparing the

detection rate of events such as Swift J0644.5−5111 with

the FRB rate.

There is an inconsistency between the inferred all-sky

γ-ray pulse rate and the FRB rate, as noted by De-

Launay et al. (2016), that also calls into question the

association. If Swift J0644.5−5111 had occurred in the

region of the BAT field of view with > 90% coding,

it would have resulted in an image-based burst trigger.

DeLaunay et al. (2016) estimate that the rate of long-

duration image-triggered events, presumably similar to

Swift J0644.5−5111, is 25 yr−1. We make the conserva-

tive assumption that these events all have FRB counter-

parts, regardless of their fluence or classification. The

100% coding region of BAT is ≈ 1000 deg2 (Barthelmy

et al. 2005), which we (conservatively) equate with

the > 90% coding region. In this region, we predict

that BAT should have been sensitive to the counter-

parts of between 8800 and 17600 FRBs in a year. We

calculate this using the fluence-complete FRB rate of

2500 sky−1 day−1 events with fluences > 2 Jy ms (Keane

& Petroff 2015), and assume both that the FRB source

counts are consistent with a Euclidean universe, and

that Swift obtains a > 50% observing duty cycle. Thus,

the odds ratio of DeLaunay et al. (2016) observing their

counterpart is the ratio of the γ-ray event rate to the

radio event rate. This places the odds ratio at between

approximately 350:1 and 700:1.

Therefore, the odds ratio of FRB 131104 having a

γ-ray counterpart (based on the disparity of the γ-

ray pulse and FRB rates), and the odds ratio of Swift

J0644.5−5111 being associated with FRB 131104 (the

calculation presented in DeLaunay et al. 2016), are com-

parable. This demonstrates that a true association is

not significantly more likely than the probability of an

unassociated occurrence. This issue was qualitatively

acknowledged by DeLaunay et al. (2016). To reconcile

the event rates would require FRB 131104 to be of a

fundamentally different, much rarer class than the other

FRBs.

Furthermore, the false alarm probabilities of such

unassociated occurrences given by DeLaunay et al.

(2016) are likely underestimated. The calculations relied

on estimating the background rate of 4.2σ image-plane

detections, when lower-significance detections may still

have exceeded their final false-alarm probability thresh-

Fast radio burst counterparts 3

old of 3σ. Their background rate was also only calcu-

lated for events with 200 s to 400 s durations, whereas

they may still have claimed a counterpart discovery

with either a shorter or longer event coincident with

FRB 131104. The false alarm rate for all the possi-

ble associations that DeLaunay et al. (2016) could have

claimed is hence likely higher than was estimated.

3. RADIO-INTERFEROMETRIC OBSERVATIONS

Figure 1. Radio-continuum images of the field surround-ing FRB 131104 in the 5.5 GHz band (top) and 7.5 GHzband (bottom). The blue circle shows the beam of Parkestelescope (to twice the half power point, which is approxi-mately the first null in the beam pattern). The 5.5 GHz and7.5 GHz image rms flux densities within the blue circle are15 µJy beam−1 and 20 µJy beam−1 respectively. The red cir-cle shows the 90% confidence region for the Swift transient.The black circle shows the position of the unusual variableAT J0642.9−5118. In both plots, the grayscale ranges lin-early from −100 to 500 µJy beam−1.

We commenced monitoring the field of FRB 131104

with the Australia Telescope Compact Array (ATCA)

3 d after the FRB was detected at Parkes. Our observa-

tions were conducted over 25 epochs spanning 2.5 yr.

Visibilities were computed using the Compact Array

Broadband Backend (Wilson et al. 2011) over two 2 GHz

width tuneable bands, centered at 5.5 and 7.5 GHz. A

42-pointing mosaic was necessary to cover to twice the

half-power beam point of Parkes observations (which is

the first null in the primary beam) at the highest fre-

quency of the ATCA observations. This was especially

crucial because of the possibility of a population of ul-

trabright FRBs that could be detected in the outer main

beam or sidelobe of the telescope (Vedantham et al.

2016b).

Observations were conducted in a variety of array con-

figurations, with maximum baseline lengths varying be-

tween 214 m and 6 km. Usually 6 antennas were avail-

able, but some observations were conducted with 5 an-

tennas (particularly in the lower resolution arrays where

inclusion of a sixth very distant antenna complicates

imaging), and one with 4. The lower spatial resolution

observations suffered from higher noise, but other prob-

lems such as source confusion were not a problem be-

cause the field is relatively sparse. Data were bandpass

calibrated using observations of either PKS 0823−500

or PKS 1934−638, and flux calibrated using the latter.

Phase calibration was conducted with regular observa-

tions of the unresolved radio galaxy J0625−5438. Data

were reduced using the miriad data reduction pack-

age (Sault et al. 1995). The visibilities for each point-

ing were imaged and deconvolved independently (using

multi-frequency synthesis and cleaning) and then com-

bined to form a composite image. Noise levels were typi-

cally 30 µJy beam−1 in the mosaicked observations. We

investigated the role of self calibration (both phase-only

and amplitude-and-phase self calibration) on our flux-

density measurements. We found that while self cali-

bration improved image fidelity it did not significantly

alter flux-density measurements1.

Figure 1 shows composite images formed from the

5.5 GHz (top) and 7.5 GHz (bottom) observations of

the field. The rms noise levels in the two images are,

respectively, 14 and 19 µJy beam−1. The width of the

primary beam of Parkes, to the first null, is shown as

the blue circle. The 90% containment region for Swift

J0644.5−5111 is shown as the red circle. There are no

sources within this region in either the mosaics of all our

data shown in Figure 1 or in individual epochs, allowing

us to place 5σ limits on persistent sources at 5.5 GHz

and 7.5 GHz of 70 and 100 µJy respectively.

1 Johnston et al. (2016) noted a ≈ 10% downward bias in flux-density measurements in mosaicked data sets. We found thatthis was mitigated by imaging with the source of interest at thereference pixel of the image.

4 R. M. Shannon & V. Ravi

Within the field of view, we have identified a strongly

variable source, which we refer to as AT J0642.9−5118.

The location of the source on the sky is (J2000) α =

6h42m57s.154(3), δ = −51◦18′17”.70(7). The light curve

for the source is presented in Figure 2. In the week af-

ter the occurrence of FRB 131104, the source brightens

by a factor of 2, exceeding 1200 µJy in the 7.5 GHz

band. During the brightening, the spectrum also in-

verts. Other sources in the field do not show this level

of variability, suggesting that mis-calibration has not in-

troduced the flux variation.

After identifying AT J0642.9−5118, we conducted

more sensitive single-pointing observations at 2.1, 5.5,

and 7.5 GHz. Observation and data reduction in the

2.1 GHz band followed the same procedures as in the

mosaicked observations. In the 2.1 GHz band, the ma-

jor differences were that only 2 GHz of bandwidth was

available, and phase calibration used the radio galaxy

PKS 0647−475. For these targeted observations, image

rms noise was typically 30 µJy beam−1 in the 2.1 GHz

band, and 10 µJy beam−1 in the higher-frequency im-

ages.

There was a modest re-brightening of

AT J0642.9−5118 approximately 300 d after the

initial flare. After the initial flare the flux density at

7.5 GHz has a mean value of 395 µJy and an rms value

of 80 µJy, suggesting that the 1.2 mJy event is a 10σ

event temporally coincident with the FRB; at 5.5 GHz

the mean flux density has been 390 µJy with an rms

value 100 µJy after the flare. The modulation index in

the 5.5 GHz (7.5 GHz) band is 0.3 (0.2) when excluding

the first three observations and 0.4 (0.5) when including

them.

4. OPTICAL OBSERVATIONS OF AT J0642.9−5118

4.1. Imaging

Following the identification of AT J0642.9−5118, we

obtained images of its optical counterpart with the

Gemini Multi-Object Spectrograph (GMOS; Hook et al.

2004) at the Gemini-South telescope. Our observations

were conducted on MJD 57141 in the g′ and r′ bands us-

ing the Hamamatsu CCDs (Gimeno et al. 2016), under

photometric conditions with 0.6′′ FWHM seeing. Four

dithered exposures were taken in each band, totalling

2617 s in the g′-band and 2322 s in the r′-band. We re-

duced the data using the standard GMOS pipeline tasks

in the Gemini IRAF package. We used facility bias and

flat-field exposures nearest in time to our observations

to correct the data, and co-added all images following

subtraction of dithering offsets. Astrometric corrections

were applied to the images using D. Perley’s autoastrom-

Figure 2. Radio light curve for the variable radio sourceAT J0642.9−5118. From the lower to upper panels weshow the flux density measured in the bands centered at2.1 GHz, 5.5 GHz, and 7.5 GHz. The x-axis is the time sinceFRB 131104. For ∆t < 100 d we show the light curve on alogarithmic time axis. For ∆t > 100 d the axis is linear.

etry software2, using the USNO B1.0 catalog as a refer-

ence (Monet et al. 2003), with 0.32′′ accuracy.

We identified a point-like counterpart to the radio

source that is the north-west component of a close

(0.6′′ separation) double (Fig. 3, left and middle panels).

We term this source G1, and its south-eastern compan-

ion G2. As we did not observe a photometric standard

field, we used the GMOS-South photometric equation

defined online3 to set the flux scale. We modeled thepoint-spread function using nearby stars and used this

to model G1 and G2 as two point sources, finding a sat-

isfactory fit to the observation. For G1, we obtained AB

magnitudes of g′ = 22.82 ± 0.02 and r′ = 22.51 ± 0.02,

and for G2 we obtained g′ = 22.87 ± 0.02 and r′ =

21.77± 0.01. At this position, the Galactic extinction is

0.208 magnitudes in the g′ band, and 0.144 magnitudes

in the r′ band (Schlafly & Finkbeiner 2011).

We also obtained imaging observations in the J band

with the FourStar instrument (Persson et al. 2013) on

the Magellan-Baade telescope at Las Campanas Obser-

vatory. The observations, conducted on MJD 57270 un-

der photometric conditions with 0.65′′ FWHM seeing,

2 http://www.astro.caltech.edu/~dperley/programs/autoastrometry.py

3 https://www.gemini.edu/sciops/instruments/gmos/calibration/photometric-stds

Fast radio burst counterparts 5

were split into 18 dithered exposures totalling 1153 s.

The data were reduced using the standard FourStar

pipeline. We calibrated the photometry and astrome-

try of the image using 2MASS point sources (Skrutskie

et al. 2006), attaining 0.2′′ astrometric accuracy. The re-

sulting detections of G1 and G2 are shown in the right

panel of Fig. 3. Using the same technique as above, we

measure AB magnitudes of J = 21.54±0.03 for G1, and

J = 20.24 ± 0.01 for G2. The Galactic extinction in

the J band is 0.045 magnitudes (Schlafly & Finkbeiner

2011).

The point source catalog of the Widefield Infrared

Survey Explorer (WISE; Wright et al. 2010) contains a

source, WISE J064257.16-511817.8, which is coincident

with G1 and is detected in the two shortest wavelength

bands. Its (AB) magnitudes are W1 = 17.9 ± 0.1 and

W2 = 16.7 ± 0.2. Based on this color, the source is

consistent with an active galactic nucleus (AGN; Stern

et al. 2012). The optical colors are also consistent with

an AGN at moderate redshift, such that Lyα is blue-

ward of our observations (Smolcic et al. 2006).

4.2. Spectroscopy

We also obtained optical spectra of G1 and G2 using

GMOS at Gemini-South. We used a 1′′ longslit oriented

along the axis of G1 and G2 (position angle of 317◦).

Our observations in the red part of the spectrum were

conducted on MJD 57362 using the R400 grating with

the GG455 order-blocking filter. Four 920 s exposures

were taken at a mean airmass of 1.2, with two centered

on 8610 A and two centered on 8510 A to cover the gaps

between CCDs. Our blue observations were conducted

on MJD 57367 at a mean airmass of 1.1 using the B600

grating with no filter, and three 920 s exposures (two

centered on 5060 A, and one centered on 4960 A). We

reduced the data using the standard GMOS pipeline,

with a bias observation obtained on MJD 57363, and

flat-field observations taken in between our science ex-

posures.

Unfortunately, intermittent high cirrus was present,

precluding accurate spectrophotometric calibration and

making sky emission lines difficult to subtract. We

nonetheless used observations of a spectrophotometric

calibrator on MJD 57562 (LTT 3218) to calibrate tel-

luric absorption features. As the seeing on both nights

was ∼ 1′′ FWHM, as measured from acquisition im-

ages, we could not deconvolve the spectral traces of G1

and G2. Furthermore, some light from G1 was likely

refracted out of the slit as GMOS does not contain an

atmospheric dispersion corrector. Hence, following op-

timal extraction of the spectra, we only considered data

taken using the R400 grating at wavelengths shorter

than 9250 A and divided the data by a smooth poly-

nomial fit to the continuum. The resulting spectrum,

which contains numerous artifacts due to imperfect sky

subtraction and is binned to 8 A resolution, is shown in

Fig. 3.

We tentatively identify two redshifted systems in this

spectrum: one at z = 0.805 ± 0.001 and one at z =

0.8875 ± 0.001. The first system is consistent with the

spectra of early-type galaxies (Polletta et al. 2007), a

hypothesis which is additionally evidenced by a possi-

ble spectral break in the continuum around 4000 A. We

thus interpret it as corresponding to G2. The redshift

of the second system, presumably G1 (the radio source),

is estimated primarily using the strong emission line

at 7040 A and by assuming (based on the WISE and

blue optical colors) that it is an AGN. Identifying the

7040 A line with [OII]3272 A results in a clear predic-

tion, specifically that strong emission lines (e.g., Lyα,

CIV, MgII) should be seen at shorter wavelengths. It

appears to exhibit Hγ and Hβ in absorption. The lack

of these normally broad lines in emission, combined with

the compact nature of its radio counterpart evidenced by

the variability, is suggestive of a narrow-line radio galaxy

(Osterbrock 1978), or perhaps a radio-loud narrow-line

Seyfert 1 (Komossa 2008). More sensitive spectra with

broader wavelength coverage would help in this classi-

fication, for example by searching for the FeII emission

features that distinguish the narrow-line Seyfert 1 class.

5. DISCUSSION

5.1. Limits on afterglows from the γ-ray pulse

Our observations can be used to search for afterglow

emission associated with the potential γ-ray transient

Swift J0644.5−5111. In the classic fireball model (Frail

et al. 2000), the flux density of radio synchrotron emis-

sion is directly related to the input energy. Murase

et al. (2016) calculated the flux density assuming thespectrum is not self absorbed and that the frequency of

interest is below the peak of the spectrum so that the

flux density is still rising. This is a reasonable assump-

tion for our observations within 9 and 10 d after the

FRB. Assuming a distance D = 3.3 Gpc consistent with

the pulses extragalactic dispersion measure, after time

T = 10 d at a frequency ν = 5.5 GHz, the flux density

is

Sν = 470 µJy( ν

5.5 GHz

)1/3(

Eγ1051.7 erg

)5/6

×( n

0.1 cm−3

)1/2 ( εB10−2

)1/3 ( εe0.1

)−2/3f5/3e

× g(2.4)−2/3(

T

10 d

)1/2 (D

3 Gpc

)−2, (1)

where E is the energy emitted in γ-rays, n is the elec-

tron number-density of the shocked medium, εB is the

magnetic energy fraction, εe is the nonthermal-electron

6 R. M. Shannon & V. Ravi

21012

2

1

0

1

2

DEC

off

set

(arc

sec)

g'-band

21012

RA offset (arcsec)

2

1

0

1

2

r'-band

21012

2

1

0

1

2

J-band

Figure 3. Upper: images of optical counterpart and its pair in g′, r′ (GMOS-South), and J bands (Magellan). Source G1,which is the counterpart to AT J0642.9−5118, is in the image center, and G2 is to the lower left (south east). The intensityscaling (zero-point) is common between the images. Bottom: Continuum-divided combined spectrum obtained of the pair ofoptical sources. We show features potentially associated with source G1 in black and source G2 in red, with the correspondingredshifts quoted in the figure. Note that numerous artefacts due to imperfect sky-line subtraction are present. The instrumentalFWHM of the spectrograph was 3 A.

energy density, fe is the nonthermal electron energy, and

g(s) = (s−2)/(s−1). For fiducial assumptions for these

parameters, we could have detected the source in both

the 5.5 GHz and 7.5 GHz observations 10 d after the

explosion, with a significance of 12 − 20σ. This sug-

gests that either the input luminosity is smaller than

estimated in DeLaunay et al. (2016), that the environ-

ment surrounding the burst is unlike that of long γ-ray

bursts or core-collapse supernova explosions, or that the

γ-ray transient is unrelated to the FRB or spurious.

5.2. The ATCA variable source

We interpret the variable radio source

AT J0642.9−5118 as emission from compact com-

ponents in a radio-loud AGN. This is evidenced by

the persistent radio variability on timescales of days to

months, the optical to mid-infrared colors of its host

system, and its possible spectral identification. The

lightcurve of the flare following FRB 131104, with the

spectrum inverting when it brightens, is consistent

with the classic picture of an expanding and coolingsynchrotron bubble. Although the flare lightcurve

is consistent with the radio afterglows of relativistic

transients (e.g., Frail et al. 2000), the persistence and

low-level variability of the radio source beyond the flare

means that we have no evidence to favor a transient

coinciding with a variable radio source, over simply a

variable radio source.

AT J0642.9−5118 is nonetheless interesting. This ob-

ject has substantial differences from the variable radio

source identified with FRB 150418 (Keane et al. 2016).

First, we clearly identify the flare of AT J0642.9−5118

with the days immediately after the FRB, as we observe

the flux density rise and the spectrum invert. Addition-

ally, scintillation in the Milky Way interstellar medium

is less likely to cause the variability of AT J0642.9−5118

because the source is at a relatively high galactic lati-

tude (b = −22◦), and the scattering is expected to be

in the weak regime (e.g., Walker 1998) at 7.5 GHz. It

Fast radio burst counterparts 7

is possible that the source was magnified by an extreme

scattering event (Bannister et al. 2016), but that is im-

probable as only one in ≈ 2000 compact sources are un-

dergoing one at a given time. Perhaps most importantly,

AT J0642.9−5118 has not re-brightened to within a fac-

tor of two of its flux densities as the peak of the flare, un-

like the case for FRB 150418 (Williams & Berger 2016;

Johnston et al. 2016). The flare of AT J0642.9−5118

following FRB 131104 thus appears to be a transient

occurrence within the scope of our monitoring of its flux

density.

Blind surveys for transients at lower frequencies find

objects with such extreme variability (factor of three

on few-month timescales) only very rarely (e.g., Moo-

ley et al. 2016). This is not surprising. Assuming con-

stant brightness temperature, intrinsic AGN variability

timescales scale with frequency proportional to ≈ ν−1.

Transient AGN flare events at higher frequencies are

hence generally expected to be shorter in time, and are

often also larger in modulation, than at lower frequen-

cies (e.g., Hovatta et al. 2008). Scintillation timescales

in the strong scintillation regime are expected to be

more rapid. The post-FRB flare of AT J0642.9−5118

is clearly most dramatic at the highest observing fre-

quency.

The temporal coincidence of AT J0642.9−5118 flare

with FRB 131104 nonetheless motivates us to consider

the possibility that it is associated with the FRB. In

this case, AGN activity would be implicated in FRB

production. The potential redshift (z = 0.8875) of

AT J0642.9−5118 is consistent with the extragalactic

DM of the FRB (Dolag et al. 2015). While other source

channels have been more strongly advocated for FRBs,

it is not implausible that AGN could produce FRBs.

Millisecond-duration radio pulses propagating through

relativistic plasma in AGN jets may be immune to both

absorption and scattering effects (Lyubarsky 2008), im-

plying that FRBs originating close to launching regions

could be observed from AGN viewed along the jets.

Mechanisms (e.g., Romero et al. 2016) have been pro-

posed for the production of FRBs in AGN jets, analo-

gous to the mechanisms for generating TeV photons.

There are however reasons to disfavor an association

between AT J0642.9−5118 and the FRB. The back-

ground transient and variable event rate at 7.5 GHz

(where the flare is the most prominent), and hence

the false-alarm rate for the association, is poorly con-

strained. Even so, the FRB rate needs to be reconciled

with the background rate (Vedantham et al. 2016a). In-

trinsic AGN variability is likely to dominate the back-

ground slow-transient rate. A detailed analysis of the

radio AGN population and its variability properties in

comparison with the FRB rate would be required to as-

sess how commonly a single object would be expected

to emit an FRB, and what its signature could be. Fur-

ther physical modeling of the conditions and orienta-

tions under which FRBs could escape AGN would help

refine such an analysis. This analysis would be further

aided by a large area survey for transient and variable

sources at high frequencies, as well as dedicated follow-

up observations of other FRBs, in particular to assess

the frequency of short-duration flares in AGN. However,

a substantially more constraining result would be the di-

rect interferometric localization of a population of FRBs

to flaring AGN.

6. CONCLUSIONS

We present 25 epochs of centimetric imaging obser-

vations of the field of FRB 131104 with the Australia

Telescope Compact Array spanning 2.5 yr. No radio af-

terglow is coincident with the γ-ray event reported by

DeLaunay et al. (2016). This tightly constrains the en-

ergetics the associated cataclysm, or suggests, as sup-

ported by probabilistic arguments we outline, that the

γ-ray event is unrelated to the FRB or spurious.

We have identified an unusual flaring radio source

temporally and spatially coincident with FRB 131104.

This source, AT J0642.9−5118, is not spatially co-

incident with the potential γ-ray transient Swift

J0644.5−5111. AT J0642.9−5118 is consistent with

compact emission components in an AGN, as identified

by optical and infrared photometry and spectroscopy.

The discovery of further, better-localized FRBs with ei-

ther radio or γ-ray flares (or neither) will resolve the

uncertainty (or not) in the multiwavelength associations

with the enigmatic fast radio burst population.

We thank M. Kasliwal for obtaining and reducing the

J-band Magellan data presented in this paper, H. Vedan-

tham, S. Kulkarni, K. Masui, R. Blandford, and S. John-

ston for useful discussions, and S. Ryder and the In-

ternational Telescope Support Office at the Australian

Astronomical Observatory for assistance in coordinat-

ing Gemini observations. We thank the group of M.

Bailes at the Swinburne University of Technology for

making available their real-time FRB detector, without

which the rapid follow-up observations of FRB 131104

would not have been possible. We are also grateful for

the prompt scheduling of our observations by CSIRO

Astronomy and Space Science operations staff. The

Australia Telescope Compact Array and Parkes radio

telescope are part of the Australia Telescope which is

funded by the Commonwealth of Australia for opera-

tion as a National Facility managed by the Common-

wealth Science and Industrial Research Organization

(CSIRO). This paper is partially based on observations

obtained at the Gemini Observatory, which is operatedby the Association of Universities for Research in As-

8 R. M. Shannon & V. Ravi

tronomy, Inc., under a cooperative agreement with the

NSF on behalf of the Gemini partnership: the National

Science Foundation (United States), the National Re-

search Council (Canada), CONICYT (Chile), Ministe-

rio de Ciencia, Tecnologıa e Innovacion Productiva (Ar-

gentina), Ministerio da Ciencia, Tecnologia e Inovacao

(Brazil), and, previously, the Department of Industry

and Science (Australia). This paper includes data gath-

ered with the 6.5 meter Magellan Telescopes located at

Las Campanas Observatory, Chile.

Facility: Parkes, ATCA, Gemini:South (GMOS),

Magellan:Baade (FourStar)

REFERENCES

Akiyama, K., & Johnson, M. D. 2016, ApJL, 824, L3

Bannister, K. W., Stevens, J., Tuntsov, A. V., et al. 2016,

Science, 351, 354Barthelmy, S. D., Barbier, L. M., Cummings, J. R., et al. 2005,

SSRv, 120, 143

Cai, Y.-F., Sabancilar, E., Steer, D. A., & Vachaspati, T. 2012,PhRvD, 86, 043521

Cordes, J. M., & Wasserman, I. 2016, MNRAS, 457, 232DeLaunay, J. J., Fox, D. B., Murase, K., et al. 2016, ApJL, 831,

L1

Dolag, K., Gaensler, B. M., Beck, A. M., & Beck, M. C. 2015,MNRAS, 451, 4277

Frail, D. A., Waxman, E., & Kulkarni, S. R. 2000, ApJ, 537, 191

Gehrels, N., Ramirez-Ruiz, E., & Fox, D. B. 2009, ARAA, 47,567

Gimeno, G., Roth, K., Chiboucas, K., & et al. 2016, Proc. SPIE,

9908, 99082SHook, I. M., Jørgensen, I., Allington-Smith, J. R., et al. 2004,

PASP, 116, 425

Hovatta, T., Nieppola, E., Tornikoski, M., et al. 2008, A&A, 485,51

Johnston, S., Keane, E. F., Bhandari, S., et al. 2016, ArXive-prints, arXiv:1610.09043

Keane, E. F., & Petroff, E. 2015, MNRAS, 447, 2852

Keane, E. F., Johnston, S., Bhandari, S., et al. 2016, Nature,530, 453

Komossa, S. 2008, in Revista Mexicana de Astronomia y

Astrofisica Conference Series, Vol. 32, Revista Mexicana deAstronomia y Astrofisica Conference Series, 86–92

Lorimer, D. R., Bailes, M., McLaughlin, M. A., Narkevic, D. J.,

& Crawford, F. 2007, Science, 318, 777Lyubarsky, Y. 2008, ApJ, 682, 1443

Masui, K., Lin, H.-H., Sievers, J., et al. 2015, Nature, 528, 523

Monet, D. G., Levine, S. E., Canzian, B., et al. 2003, AJ, 125,984

Mooley, K. P., Hallinan, G., Bourke, S., et al. 2016, ApJ, 818,105

Murase, K., Meszaros, P., & Fox, D. B. 2016, ArXiv e-prints,

arXiv:1611.03848

Osterbrock, D. E. 1978, PhyS, 17, 137

Persson, S. E., Murphy, D. C., Smee, S., et al. 2013, PASP, 125,

654

Polletta, M., Tajer, M., Maraschi, L., et al. 2007, ApJ, 663, 81

Ravi, V., Shannon, R. M., & Jameson, A. 2015, ApJL, 799, L5

Ravi, V., Shannon, R. M., Bailes, M., et al. 2016, Science, in

press

Romero, G. E., del Valle, M. V., & Vieyro, F. L. 2016, PhRvD,

93, 023001

Sault, R. J., Teuben, P. J., & Wright, M. C. H. 1995, in

Astronomical Society of the Pacific Conference Series, Vol. 77,

Astronomical Data Analysis Software and Systems IV, ed.

R. A. Shaw, H. E. Payne, & J. J. E. Hayes, 433

Schlafly, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103

Skrutskie, M. F., Cutri, R. M., Stiening, R., et al. 2006, AJ, 131,

1163

Smolcic, V., Ivezic, Z., Gacesa, M., et al. 2006, MNRAS, 371, 121

Spitler, L. G., Scholz, P., Hessels, J. W. T., et al. 2016, Nature,

531, 202

Stern, D., Assef, R. J., Benford, D. J., et al. 2012, ApJ, 753, 30

Thornton, D., Stappers, B., Bailes, M., et al. 2013, Science, 341,

53

Vedantham, H. K., Ravi, V., Mooley, K., et al. 2016a, ApJL,

824, L9

Vedantham, H. K., Ravi, V., Hallinan, G., & Shannon, R. M.

2016b, ApJ, 830, 75

Walker, M. A. 1998, MNRAS, 294, 307

Williams, P. K. G., & Berger, E. 2016, ApJL, 821, L22

Wilson, W. E., Ferris, R. H., Axtens, P., et al. 2011, MNRAS,

416, 832

Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al. 2010,

AJ, 140, 1868